How to Draft Churn-Save Offers
A practical guide to churn-save replies with an agent: reason classification, matched offers, a written policy, and approval before anything is sent.
Founder, Task Machine
A churn-save offer is the reply you send when a customer tells you they are cancelling: a message that acknowledges their stated reason, makes exactly one offer matched to that reason, and asks one honest question about what went wrong. It targets voluntary churn, the customers who decided to leave, at the one moment they are still listening.
It is among the highest-leverage messages a subscription business sends. The customer has already told you the problem in their cancellation note. A reply that addresses that specific problem can keep the account, and even when it cannot, the exit answer tells you what to fix next.
Why unanswered cancellations quietly cost you
A cancellation note is a closing window. The customer wrote it while the decision was fresh, and every day it sits unanswered, the decision hardens. Most small teams respond in one of two ways: a template that says "sorry to see you go", or a blanket discount sent to everyone regardless of reason.
Both waste the moment. A discount cannot save someone who stopped using the product, and a feature roadmap cannot save someone who can no longer afford it. The blanket discount is the more expensive mistake, because it teaches customers that cancelling is how you get a deal. And when nobody owns the reply, the notes that arrive in a busy week get the template or nothing at all.
What the manual process looks like
Done by hand, a churn save is a short ritual that rewards care over speed:
- Read the cancellation note closely and work out the real reason, which is not always the stated one.
- Classify it: too expensive, not using it enough, missing a feature, switching to a competitor, technical issues, a temporary pause in need, or a business that closed.
- Pick the one offer that fits the reason, from the set of concessions you are actually willing to give.
- Write the reply: genuine thanks, a specific acknowledgement of their reason, the offer with its concrete benefit, and one exit-interview question.
- Check the draft against your own limits, such as the discount cap and whether this customer has already had an offer, then send it.
Each note takes maybe twenty minutes to do well. The work is not hard, but it demands the same discipline every time, and the discipline is exactly what slips when cancellations cluster.
What an agent can automate
Everything in that ritual except the final send decision is a rule that can be written down, which makes it a good fit for an agent running a fixed workflow:
- Read and classify. The agent reads the cancellation note or attachment and extracts the stated and implied reason, the tone, the customer's value, and whether the situation is recoverable, then classifies the reason into one bucket.
- Match exactly one offer. Each reason maps to a different save. Too expensive gets a time-boxed discount or a cheaper tier, with the dollar amount saved spelled out. Not using it gets a pause of one to three months plus an onboarding nudge, never a discount. A missing feature gets an honest roadmap answer or a workaround, with no ship date that is not committed. A switch to a competitor gets empathy and one honest differentiator. Technical issues get an escalation to support instead of a deal. A closed business gets a thank-you and a graceful goodbye.
- Draft the reply. The draft opens with genuine empathy, acknowledges the reason specifically enough to show the note was actually read, presents the one matched offer with its concrete benefit, and asks one exit-interview question in "help us improve" framing. It also makes leaving easy regardless, because a save attempt that traps people saves nothing.
- Self-check against policy. Before the draft reaches you, the agent checks it against every hard limit in the save-offer policy: no discount beyond 30% off, every offer time-boxed, never more than one offer per customer, no fake scarcity, and no claim the business cannot honor.
What stays judgment: whether the message goes out at all, and every case the rules cannot settle. High-value accounts, serious bug complaints, ambiguous notes, and any save that would need an out-of-policy offer get escalated to you instead of guessed at.
The guardrails that make it safe
A save offer is a pricing decision sent to a customer who is already halfway out the door, which is exactly the kind of message that should wait for a person.
The first guardrail is a written policy. The offer limits live in a save-offer policy document, not in the agent's head, so what a draft may offer is something you can read, edit, and tighten. The second is the workflow itself: intake, offer matching, drafting, and the policy self-check all run before an explicit approval step, and nothing is sent without your sign-off. Only after you approve does the agent prepare the reply in your email or support tool through the browser. The escalation rules close the loop, routing top accounts and edge cases to a personal reply instead of a template.
Set it up in Task Machine
The Churn-save drafter playbook installs everything above as working records in your workspace: the Retention Agent carrying the reason-matching method, the drafting workflow with the approval step built in, the save-offer policy document in a Retention folder, and the two skills covering churn prevention and offer design. Setup takes a few minutes. You need a Task Machine workspace and permission to install playbooks (workspace owners have it). Connecting your email or support tool is not required up front; until you authorize browser access, the agent drafts into the Retention folder from the cancellation note you paste in and the save-offer policy.
1. Find the playbook
Open Playbooks in your workspace and search for "churn-save", or browse to the Sales category. The card lists what the playbook creates and the models its agent runs on.
2. Preview what it installs
Preview & install opens the full contents before anything is created: the Retention Agent, the Churn-save drafter workflow, the save-offer policy document, the two skills carrying the retention method, and the payment providers you can pick from. The provider entries are optional.
3. Pick your payment providers
Start setup asks for the details the agent needs. The first is the payment providers your customers bill through: Stripe, Paddle, Polar, or PayPal. Picking one lets the agent look up a customer's plan, billing history, and subscription status before drafting the save offer. The pick is optional, and only the providers you pick are installed; the others never touch your workspace.
4. Give the drafter your context and limits
Four more answers shape every draft. Customer segment tells the agent who the replies address. Churn trigger or complaint describes the cancellation reason you see most, which primes the classification. Approved save offers lists the concessions a draft may propose, and nothing outside that list appears in a reply. Tone sets how the replies sound.
5. Generate and review
Generate customized playbook bakes your answers into the agent instructions, the workflow prompts, and the policy document. The result comes back for review before anything is created. Read through the agent and workflow cards, confirm the approved offers match what you would actually give, and check that only the providers you picked appear as connected services.
6. Install
Install customized playbook creates everything in one step and lists what landed in your workspace. Two follow-ups arrive in your inbox: Set churn-save offer guardrails, which opens the save-offer policy so you can define allowed discounts, extension rules, escalation cases, refund boundaries, and off-limits promises, and Start Churn-save drafter, which walks you through the workflow before you use it on a real customer. From then on, every cancellation note you drop into the workflow runs the full loop, and the finished draft waits in your inbox for approval before anything reaches the customer.
What good looks like
Three signals tell you whether the process works:
- Every offer traces to the reason. Reading a week of drafts, you should be able to name the reason bucket from the offer alone. A run of discounts going to customers who stopped logging in means the matching has drifted.
- Pauses come back. When a pause is the right match, it usually is for customers whose need is temporary: 60 to 80% of customers who pause return. A growing share of pauses over discounts is a healthy sign.
- No customer hears an offer twice. The once-per-customer limit is the difference between a save program and a discount hotline. Any repeat offer in the log is a policy breach worth investigating.
Common questions
Should every cancelling customer get a discount? No. A discount only fixes a price objection. Sent to everyone, it trains customers to cancel whenever they want a deal, and it does nothing for the customer who left because of a bug or a missing feature. Match the offer to the reason and reserve discounts for "too expensive".
How big should a save discount be? 20 to 30% off for two to three months is the working range, always time-boxed, never beyond 30%, and stated as the dollar amount saved rather than a percentage. Deeper cuts save the account this month and devalue the product permanently.
What if the customer's reason is a bug or an outage? No deal. A discount on a broken product reads as an insult. The right move is an immediate escalation to support, optionally with a credit, and that is exactly what the agent does with that reason bucket.
Does a save attempt make cancelling harder? It should not. Every draft makes leaving easy regardless of whether the customer takes the offer, uses no fake scarcity or countdown, and closes with a genuine thank-you and an easy reactivation path. The save is the offer, never friction.
Can this run before connecting an email or support tool? Yes. Paste the cancellation note into the workflow and the agent drafts the reply into the Retention folder, working from the note and the save-offer policy. Authorizing browser access to your email or support tool lets it prepare the approved reply where you actually send from.